STAGE 4: FIRST DATA COLLECTION AND ITEM ANALYSIS

After conducting the content validity evaluation, the researcher retained a set of items that have been carefully invented and reviewed. The retained items were then presented to an appropriate sample through online survey platform Amazon MTurk for the first data collection with which EFA was conducted. The objective was to examine how well those items represent the psychometric properties of the new scale and to reduce the initial items to a smaller and more parsimonious set of variables. Initial reliability of the scale was also evaluated with several reliability measures.

3.5.1 PARTICIPANTS

The population of the current study was sports consumers who attend live sporting events. Thus, anyone who has a history of attending live sporting events was entitled to participate in the study. Although the screening and selecting qualified

participants were not very restrictive, potential respondents still had to clearly remember the events they attended in order to evaluate the developed sensory items based on their experiences. Therefore, the researcher selected respondents who visited any live sporting events within the last three months from the survey participation. The three month period might be too long, but the decision of which participants should be involved in the

development of a scale depends upon the feasibility of recruiting qualified respondents (Spector, 1992). The researcher had a concern that if the period was shortened to less than a month, it might not have been possible to recruit enough number of participants in a timely manner. In addition, the three reminder questions- 1) What was the live

professional sporting event you attended most recently? 2) When was the live

professional sporting event you attended most recently? and 3) What was the location of the live professional sporting event you attended most recently?- forced respondents to retrace their memory. Another delimitation of the sampling strategy was age. Participants whose age was younger than 18 were filtered out because the researcher envisioned to select respondents who have enough experience in live sporting events and who

independently attend games. No other restriction was applied, such as types of sports or geographic locations with the intention of maximizing external validity of the scale.

A total of 185 sports consumers answered the questionnaire regarding their experiences at a variety of live sporting events ranging from the four major leagues to the Ultimate Fighting Championships (UFC). Forty-one percent of the respondents stated that they attended the live event within one month from the survey participation and thirty-two percent reported the event was within two months. Over half of the

respondents were between 30 and 49 years of age (55.1%) and other age groups were 35% of 18 to 29, 7% of 50 to 69, and 1% of 70 or older. Fifty-six percent of the sample was male, and the racial breakdown of the participants was 70% Caucasian, 12% African American, 7% Hispanic, 9% Asian, and 2% self-identified Others. Overall, demographics of the sample were not highly discrepant with the population of sports fans in the United States (“Demographics of Sports Fans,” 2017).

3.5.2 SAMPE SIZE

There has been considerable discussion over the sample size needed to carry out tests of statistical significance in EFA. Some scholars recommend an exact number or range of participants. For example, Nunnally (1978) recommends 300 as an adequate number of respondents for an EFA. Spector (1992) suggests a range of 100 to 200

participants. Other researchers determine the sample size needed to achieve robust results based on the number items. For example, Rummel (1970) recommends the item-to- response ratio of 1:4 and Schwab (1980) suggests the ratio of 1:10 for a scales to be factor analyzed. There is no generally accepted “rule of thumb,” and strict rules concerning sample size for EFA have mostly disappeared. Much research has

demonstrated that adequate sample size is mainly determined by the characteristics of the data (MacCallum, Widaman, Zhang, & Hong, 1999). As long as item correlations are reasonably high enough, a small sample can be used for an accurate analysis. One study revealed that a sample size of 150 observations is sufficient in most cases to obtain a robust solution in EFA (Guadagnoli & Velicer, 1988). The sample size used in the study was 185, which exceeded Guadagnoli and Velicer’s recommendation (1988) and also within the range of Spector’s suggestion (1992) of 100 to 200 participants preventing potential data analysis problems.

3.5.3 INSTRUMENT AND DATA COLLECTION PROCEDURE

An online survey was created using research software Qualtrics (see Appendix D for the survey), which automatically generated an anonymous hyperlink. The hyperlink was then distributed through an online survey platform Amazon Mechanical Mturk. A total of 185 participants completed the survey receiving $1 each as compensation. The

SIF scale was measured with the 26 items confirmed in the previous qualitative stage. Participants rated each item on a 7-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree). Two screening questions were also asked. One screening question was, “Have you attended any ticketed professional sporting event(s) within the last three months?” to check participants actually attended a sporting event. The question specified “professional sporting events” in an effort to avoid potential issues by including small or amateur events where spectators sometimes do not have enough sensory stimuli to enjoy. Participants whose age was younger than 18 was also screened out to select respondents who have enough experience in live sporting events and who independently attend games. A few demographic questions were also included at the end of the survey. Each

participant received a confirmation code at the end, which functioned as an indicator for the researcher to select or screen out respondents. The data were then exported to SPSS version 25 for subsequent analysis.

3.5.4 DATA ANALYSIS

After the data gathering, it was essential to examine whether the items adequately constitute the scale. Data examining through factor analysis is a necessary step in

determination of the viability of the scale. Two types of factor analyses are available for the scale development process: CFA and EFA. The exploratory one is typically used to reduce the items into a smaller and more parsimonious set of variables. The confirmatory type is used to evaluate the factor structure by statistically testing the significance of the model and the relationships among items and scales (Hinkin et al., 1997). Both types of analyses can be used in scale development, yet CFA is more widely used for a deductive method. (Kline, 2013). Since the researcher used an inductive approach, EFA was helpful

for identifying the structure of the scale. Moreover, focus group participants as well as the expert panel noted a concern that the taste fit might have separate dimensions of food and beverage. Therefore, EFA seemed to be a required course of action to take in the scale development process. The EFA procedure included the determination of the number of factors to retain with the usage of multiple stathstical techniques and the examination of item loadings.

The internal consistency scores of the scale items were also estimated through several reliability measures with the first data set. First, Cronbach’s alpha estimation was utilized to assess the consistency among items in each factor. Internal consistency level is considered satisfactory with coefficient alpha above .70 (Cronbach, 1951). Second, item to total statistics were measured. An acceptable item to total correlation value is .5 or greater (Hair, Black, Babin, & Anderson, 2010). Finally, inter-item correlations were examined for each factor. An acceptable inter-item correlation value should fall in the range of .3 to .8 not to hurt the reliability of the scale (Hair et al., 2010).